aboutsummaryrefslogtreecommitdiffstats
path: root/Documentation/spinlocks.txt
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 18:20:36 -0400
committerLinus Torvalds <torvalds@ppc970.osdl.org>2005-04-16 18:20:36 -0400
commit1da177e4c3f41524e886b7f1b8a0c1fc7321cac2 (patch)
tree0bba044c4ce775e45a88a51686b5d9f90697ea9d /Documentation/spinlocks.txt
Linux-2.6.12-rc2v2.6.12-rc2
Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
Diffstat (limited to 'Documentation/spinlocks.txt')
-rw-r--r--Documentation/spinlocks.txt212
1 files changed, 212 insertions, 0 deletions
diff --git a/Documentation/spinlocks.txt b/Documentation/spinlocks.txt
new file mode 100644
index 000000000000..c2122996631e
--- /dev/null
+++ b/Documentation/spinlocks.txt
@@ -0,0 +1,212 @@
1UPDATE March 21 2005 Amit Gud <gud@eth.net>
2
3Macros SPIN_LOCK_UNLOCKED and RW_LOCK_UNLOCKED are deprecated and will be
4removed soon. So for any new code dynamic initialization should be used:
5
6 spinlock_t xxx_lock;
7 rwlock_t xxx_rw_lock;
8
9 static int __init xxx_init(void)
10 {
11 spin_lock_init(&xxx_lock);
12 rw_lock_init(&xxx_rw_lock);
13 ...
14 }
15
16 module_init(xxx_init);
17
18Reasons for deprecation
19 - it hurts automatic lock validators
20 - it becomes intrusive for the realtime preemption patches
21
22Following discussion is still valid, however, with the dynamic initialization
23of spinlocks instead of static.
24
25-----------------------
26
27On Fri, 2 Jan 1998, Doug Ledford wrote:
28>
29> I'm working on making the aic7xxx driver more SMP friendly (as well as
30> importing the latest FreeBSD sequencer code to have 7895 support) and wanted
31> to get some info from you. The goal here is to make the various routines
32> SMP safe as well as UP safe during interrupts and other manipulating
33> routines. So far, I've added a spin_lock variable to things like my queue
34> structs. Now, from what I recall, there are some spin lock functions I can
35> use to lock these spin locks from other use as opposed to a (nasty)
36> save_flags(); cli(); stuff; restore_flags(); construct. Where do I find
37> these routines and go about making use of them? Do they only lock on a
38> per-processor basis or can they also lock say an interrupt routine from
39> mucking with a queue if the queue routine was manipulating it when the
40> interrupt occurred, or should I still use a cli(); based construct on that
41> one?
42
43See <asm/spinlock.h>. The basic version is:
44
45 spinlock_t xxx_lock = SPIN_LOCK_UNLOCKED;
46
47
48 unsigned long flags;
49
50 spin_lock_irqsave(&xxx_lock, flags);
51 ... critical section here ..
52 spin_unlock_irqrestore(&xxx_lock, flags);
53
54and the above is always safe. It will disable interrupts _locally_, but the
55spinlock itself will guarantee the global lock, so it will guarantee that
56there is only one thread-of-control within the region(s) protected by that
57lock.
58
59Note that it works well even under UP - the above sequence under UP
60essentially is just the same as doing a
61
62 unsigned long flags;
63
64 save_flags(flags); cli();
65 ... critical section ...
66 restore_flags(flags);
67
68so the code does _not_ need to worry about UP vs SMP issues: the spinlocks
69work correctly under both (and spinlocks are actually more efficient on
70architectures that allow doing the "save_flags + cli" in one go because I
71don't export that interface normally).
72
73NOTE NOTE NOTE! The reason the spinlock is so much faster than a global
74interrupt lock under SMP is exactly because it disables interrupts only on
75the local CPU. The spin-lock is safe only when you _also_ use the lock
76itself to do locking across CPU's, which implies that EVERYTHING that
77touches a shared variable has to agree about the spinlock they want to
78use.
79
80The above is usually pretty simple (you usually need and want only one
81spinlock for most things - using more than one spinlock can make things a
82lot more complex and even slower and is usually worth it only for
83sequences that you _know_ need to be split up: avoid it at all cost if you
84aren't sure). HOWEVER, it _does_ mean that if you have some code that does
85
86 cli();
87 .. critical section ..
88 sti();
89
90and another sequence that does
91
92 spin_lock_irqsave(flags);
93 .. critical section ..
94 spin_unlock_irqrestore(flags);
95
96then they are NOT mutually exclusive, and the critical regions can happen
97at the same time on two different CPU's. That's fine per se, but the
98critical regions had better be critical for different things (ie they
99can't stomp on each other).
100
101The above is a problem mainly if you end up mixing code - for example the
102routines in ll_rw_block() tend to use cli/sti to protect the atomicity of
103their actions, and if a driver uses spinlocks instead then you should
104think about issues like the above..
105
106This is really the only really hard part about spinlocks: once you start
107using spinlocks they tend to expand to areas you might not have noticed
108before, because you have to make sure the spinlocks correctly protect the
109shared data structures _everywhere_ they are used. The spinlocks are most
110easily added to places that are completely independent of other code (ie
111internal driver data structures that nobody else ever touches, for
112example).
113
114----
115
116Lesson 2: reader-writer spinlocks.
117
118If your data accesses have a very natural pattern where you usually tend
119to mostly read from the shared variables, the reader-writer locks
120(rw_lock) versions of the spinlocks are often nicer. They allow multiple
121readers to be in the same critical region at once, but if somebody wants
122to change the variables it has to get an exclusive write lock. The
123routines look the same as above:
124
125 rwlock_t xxx_lock = RW_LOCK_UNLOCKED;
126
127
128 unsigned long flags;
129
130 read_lock_irqsave(&xxx_lock, flags);
131 .. critical section that only reads the info ...
132 read_unlock_irqrestore(&xxx_lock, flags);
133
134 write_lock_irqsave(&xxx_lock, flags);
135 .. read and write exclusive access to the info ...
136 write_unlock_irqrestore(&xxx_lock, flags);
137
138The above kind of lock is useful for complex data structures like linked
139lists etc, especially when you know that most of the work is to just
140traverse the list searching for entries without changing the list itself,
141for example. Then you can use the read lock for that kind of list
142traversal, which allows many concurrent readers. Anything that _changes_
143the list will have to get the write lock.
144
145Note: you cannot "upgrade" a read-lock to a write-lock, so if you at _any_
146time need to do any changes (even if you don't do it every time), you have
147to get the write-lock at the very beginning. I could fairly easily add a
148primitive to create a "upgradeable" read-lock, but it hasn't been an issue
149yet. Tell me if you'd want one.
150
151----
152
153Lesson 3: spinlocks revisited.
154
155The single spin-lock primitives above are by no means the only ones. They
156are the most safe ones, and the ones that work under all circumstances,
157but partly _because_ they are safe they are also fairly slow. They are
158much faster than a generic global cli/sti pair, but slower than they'd
159need to be, because they do have to disable interrupts (which is just a
160single instruction on a x86, but it's an expensive one - and on other
161architectures it can be worse).
162
163If you have a case where you have to protect a data structure across
164several CPU's and you want to use spinlocks you can potentially use
165cheaper versions of the spinlocks. IFF you know that the spinlocks are
166never used in interrupt handlers, you can use the non-irq versions:
167
168 spin_lock(&lock);
169 ...
170 spin_unlock(&lock);
171
172(and the equivalent read-write versions too, of course). The spinlock will
173guarantee the same kind of exclusive access, and it will be much faster.
174This is useful if you know that the data in question is only ever
175manipulated from a "process context", ie no interrupts involved.
176
177The reasons you mustn't use these versions if you have interrupts that
178play with the spinlock is that you can get deadlocks:
179
180 spin_lock(&lock);
181 ...
182 <- interrupt comes in:
183 spin_lock(&lock);
184
185where an interrupt tries to lock an already locked variable. This is ok if
186the other interrupt happens on another CPU, but it is _not_ ok if the
187interrupt happens on the same CPU that already holds the lock, because the
188lock will obviously never be released (because the interrupt is waiting
189for the lock, and the lock-holder is interrupted by the interrupt and will
190not continue until the interrupt has been processed).
191
192(This is also the reason why the irq-versions of the spinlocks only need
193to disable the _local_ interrupts - it's ok to use spinlocks in interrupts
194on other CPU's, because an interrupt on another CPU doesn't interrupt the
195CPU that holds the lock, so the lock-holder can continue and eventually
196releases the lock).
197
198Note that you can be clever with read-write locks and interrupts. For
199example, if you know that the interrupt only ever gets a read-lock, then
200you can use a non-irq version of read locks everywhere - because they
201don't block on each other (and thus there is no dead-lock wrt interrupts.
202But when you do the write-lock, you have to use the irq-safe version.
203
204For an example of being clever with rw-locks, see the "waitqueue_lock"
205handling in kernel/sched.c - nothing ever _changes_ a wait-queue from
206within an interrupt, they only read the queue in order to know whom to
207wake up. So read-locks are safe (which is good: they are very common
208indeed), while write-locks need to protect themselves against interrupts.
209
210 Linus
211
212